With thanks to Kit Baum, -ranktest-, a new package by Frank Kleibergen
and Mark Schaffer, is available for download from SSC.
-ranktest- implements the Kleibergen-Paap (2006) rk test of a rank of a
matrix. Various tests in statistics and econometrics can be cast in
terms of the test of a rank of a matrix; examples are the joint
significance of regression coefficients, the rank condition in
simultaneous equations estimation, and the trace statistic in tests of
cointegration. Standard tests of the rank of a matrix, such as the
canonical correlations test, assume i.i.d. errors. The Kleibergen-Paap
rk statistic as implemented in -ranktest- allows for tests of rank that
are robust to various forms of heteroskedasticity, autocorrelation and
clustering. In the case of i.i.d. errors, the rk test simplifies to the
standard canonical correlations-based tests.
